Re: Why will we reach the singularity?

From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Sat Mar 01 2003 - 14:26:10 MST

  • Next message: Dehede011@aol.com: "Re: Communism-Capitalism 500yrs"

    Joao, now this is indeed an interesting question and I will offer
    my comments not having reviewed the other comments.

    > I've been wondering on why are transhumanists so confident that we will
    > reach the singularity.

    The question of the "singularity" (and how it is perceived) is tied up
    in the "soft" vs. "hard" takeoff scenarios. My personal opinion is
    that the "inertia" of humanity will tend to bias development towards
    the "soft" side of things (i.e. a slow ramp-up to the singularity).

    > In truth, I'm disappointed with what's being done and I want to know
    > why are transhumanists so confident we will reach the singularity.

    Some of the rest of us who have spent a great deal of their personal
    resources are disappointed as well *but* there are some interesting
    developments that offset this (the Ellison Medical Foundation comes
    to mind).

    In truth, a real "singularity" must require one of two things:
    (a) a majority of humans must accept that they have to increase
        the rate of their self-evolution and act accordingly; or
    (b) that an independent self-evolving unconstrained AI is developed.

    I don't view (b) as a good alternative for humanity unless Eliezer's
    efforts at a "friendly" AI are successful.

    We have made progress in astrophysics but it is a very slow journey.

    > It's true breakthroughs have been made in
    > biology and medicine, such as the Human Genome Project, but, shit, we
    > haven't even cured AIDS, how can we expect to cure aging anytime soon?

    Two *very* different problems (I must stress this). HIV is a problem
    involving a virus with a genome replication mechanism designed to be
    sloppy. Aging is a problem with a genome replication mechanism
    designed to be increasingly accurate (as the longevity of the
    species increases). That is not to say that accurate genome replication
    prevents aging (there are a host of other problems one has to solve
    to prevent aging) -- but that innacurate genome replication cannot
    but contribute to aging (obviously if one has evolved a genetic
    program with "minimal" aging and it becomes corrupted so that it
    is no longer that program the impact is most likely to be detrimental.)

    > Also, I'm disappointed with the way science is made in the academia with
    > personal egos rising above finding the mechanisms of aging. If we want to
    > cure aging, we need to work together, but not many do that.

    Not completely true. Since I'm not in "academia" I can function somewhat
    independently of that framework. That has allowed me positions on both
    the AGE and ExI boards of directors. So I can exert some influence
    and have done so when the opportunity has been available. Aubrey
    de Grey is doing similar functions with regards to the 10th IABG
    Congress.

    But my observations (of a decade or so of being involved in aging
    research) are that it is very much related to a lack of interest
    (belief) that the problem can be solved and therefore a lack of
    funding and/or qualified scientists going into the area (these
    are obviously related).

    > In the end, I would say that the basis for the singularity is Moore's law,
    > for it allows not only faster computers but also developments in DNA
    > sequencing and a host of other possibilities.

    Yes, obviously so.

    > Yet I'm sure there are physical limits for Moore's law. When will we
    > reach them? Can you be sure Moore's law will continue for long enough
    > to develop a smarter-than-man artificial intelligence?

    The hard limits were discussed in quite great detail in Drexler's
    Nanosystems { Sections 12.8 and 12.9 (and all of the discussion
    leading up to that) }.

    You can get 10^21 OPS, roughly 6 times greater than our best estimate
    of a brain at 10^15 OPS, but multiplied by the fact that the
    nanocomputer probably occupies a volume of ~1 cm^3 compared to the
    brain's > 1000 cm^3 (so propagation delays are significantly less
    and "intelligence" may be significantly greater).

    Or in other words -- a nanocomputer is potentially a *lot* smarter
    than humans (on the order of millions to billions of times).

    When we will reach these limits depends on to a large extent on economic
    factors. I have received information that the U.S. investment in
    venture capital declined from $90 billion to $19 billion between
    2000 and 2002. So if you want to evaluate technology and business
    development, one has to factor economics into the process. But I
    think it is safe to assume that we *will* push this envelope
    (i.e. to the limits of Moore's Law) within this century, perhaps
    within the next few decades.

    (This is, in part, Ray Kurzweil's message, but I am adapting it a bit to
    allow for my own perspectives.)

    > When I found transhumanism, already several years ago, I thought it set an
    > optimistic but plausible scenario. Now, I'm starting to wonder if we're not
    > just another cult willing to sacrifice reality towards a fairer image of
    > the world.

    We are *definitely* in the "optmistic/plausible" scenario.
    I and others have been, over the last decade, been willing to put
    "money on the line". There is clearly viewable progress
    (where in the mid-90's there were essentially two significant
    companies involved in aging research, now there are more than
    a dozen). I see progress in other transhumanistic perspectives
    as well (see recent NY Times articles on the installation of solar
    power systems at the Whitehouse).

    Joao, do not lose hope -- it is just that progress sometimes takes
    much longer than we would like (or expect).

    *But* please do not count the singularity out. It depends in large
    part how we develop its foundations.

    Robert



    This archive was generated by hypermail 2.1.5 : Sat Mar 01 2003 - 14:31:07 MST